|Publication number||US7233691 B2|
|Application number||US 10/168,811|
|Publication date||Jun 19, 2007|
|Filing date||Dec 29, 2000|
|Priority date||Dec 29, 1999|
|Also published as||CA2395257A1, CA2395257C, EP1242966A1, EP1242966A4, EP1242966B1, US20020191838, WO2001048683A1|
|Publication number||10168811, 168811, PCT/2000/35591, PCT/US/0/035591, PCT/US/0/35591, PCT/US/2000/035591, PCT/US/2000/35591, PCT/US0/035591, PCT/US0/35591, PCT/US0035591, PCT/US035591, PCT/US2000/035591, PCT/US2000/35591, PCT/US2000035591, PCT/US200035591, US 7233691 B2, US 7233691B2, US-B2-7233691, US7233691 B2, US7233691B2|
|Inventors||Jeffrey M. Setterholm|
|Original Assignee||Geospan Corporation|
|Export Citation||BiBTeX, EndNote, RefMan|
|Patent Citations (16), Non-Patent Citations (2), Referenced by (62), Classifications (10), Legal Events (8)|
|External Links: USPTO, USPTO Assignment, Espacenet|
This application claims priority from Provisional Patent Application Ser. No. 60/173,440 filed on Dec. 29, 1999, for “Passive Volume Surveying From Synthetic Quantitative Stereo Pair Images” by Jeffrey M. Setterholm.
The present invention relates to surveying from recorded images, and more particularly to passive volumetric surveying from pseudo or synthetic stereo images. Thorough understanding, precise measuring, and concise parameterization of the influences of geometric effects on the content of images results in being able to use a compact set of equations to separate the complexities of image geometry from the problem of image correlation—thus paving the way for considerably more efficient automatic image correlation as is the case with traditional coplanar stereo imagery. That, in turn, allow the efficient generation of detailed three dimensional virtual models of the real world as can be seen from the multiple arbitrary viewpoints provided by oblique imagery.
Surveying involves determining the geographic location of points or objects that appear in an image, such as a video image. The images used in surveying may be obtained in any one of a variety of ways. Examples include images taken from a helicopter and images taken from a moving vehicle. A method for obtaining video image data from a moving vehicle is disclosed in commonly assigned U.S. Pat. No. 5,633,946, entitled “METHOD AND APPARATUS FOR COLLECTING AND PROCESSING VISUAL AND SPATIAL POSITION INFORMATION FROM A MOVING PLATFORM” (the '946 patent), which is hereby incorporated by reference. The '946 patent discloses a vehicle with multiple video cameras mounted thereon at different orientations. The vehicle is driven along a street and video images from each of the video cameras are recorded. Each image frame has a time code associated therewith. Accurate spatial position data is also obtained and recorded along with associated time codes. The spatial position data is obtained from a global positioning system (GPS) and an inertial navigation system (INS). The GPS system includes a GPS receiver in the vehicle and a GPS base receiver located in a known position. Errors introduced into the GPS satellite information are eliminated through differential processing of the GPS data gathered by the GPS receiver in the vehicle and the GPS base receiver. The INS comprises a set of rotation and acceleration sensors, and continually measures changes in rotation and changes in velocity as the vehicle moves in latitude, longitude, elevation, pitch, roll and yaw. The INS, therefore, provides raw differential spatial position data with six degrees of freedom between points where precise positional readings are taken by the GPS system.
In post-processing of the above navigation data, Kalman filtering techniques are used to compute a six-degree-of-freedom trajectory for the van as a function of time. This resulting trajectory is a best-fit through all of the navigation data. The indexing of the vehicle's six-dimensional spatial position data by time code allows it to be correlated with each recorded video image, which is also indexed by time code.
For each camera, camera calibration data is also generated and stored for later use in the surveying process. The camera calibration data is intended to adequately represent both the internal optical geometry of the camera and the external six-dimensional location of the camera in the vehicle. The internal aspects relate image pixels with associated rays in camera coordinates. The external aspects relate rays defined in camera coordinates with rays in global coordinates. Both these aspects operate bi-directionally. After both vehicle navigation data post-processing and camera calibration have been completed, a six-dimensional global location can be assigned to each video frame. The combination of the spatial data and the camera calibration data is referred to herein as the image parameters for each frame. After the image parameters have been determined, an image is ready to be used for surveying. Two or more such images at different locations, which view the same object or objects of interest, are used to accomplish the surveying.
Traditionally, aerial surveying has been done using a stereo photogrammetric approach. The mathematics of traditional stereo photography are based on the assumption the source images were taken in approximately the same geometric plane with correction terms for minor deviations of image location from the coplanar assumption. Such an approach must compensate for a relatively short baseline between two cameras by using high resolution digital cameras, which require extraordinarily large amounts of data to be stored. Further, when computing the position of a point or an object which is a great distance from the cameras, as must often be done in the stereo photogrammetric approach, the potential for error in calculating the position of the object is greatly increased.
One objective of the '946 patent is to determine the location of objects seen in two or more images using a generalized approach to surveying in which the locations of objects within images are determined from any two or more images having an overlapping scene, regardless of interimage geometry. It would be desirable to systematically apply this concept to the overlapping pixels of entire images without regard to interimage geometry. However, a difficulty that arises with a plurality of non-coplanar images is that the images are not easily correlated. Correlation algorithms must search through large portions of image data to find pixels in one image that correspond to pixels in the second image. It is commonly known that correlation processes to operate more efficiently on stereo pairs images. For example, in a human-viewable stereo pair, the same object-point in the left and right images lies on the same horizontal line in both images; the difference in lateral displacement along the horizontal line varies with distance of the object-point from the co-plane of the images.
While human viewing of stereo pairs is qualitative, it is also desirable to have definitive quantitative knowledge of the depth information implicit in a stereo pair. Furthermore, automatic algorithms, not constrained by human physiology can operate on a broader class of such stereo pairs. We refer to this broader class of transformed images as Quantitative Cylispheric Stereo Pairs (QCSP'-s). QCSP'-s preserve the line registration of corresponding pixels.
In the present invention, recorded images having an overlapping scene are transformed using a geometric transformation, into QCSPs. The QCSPs can take a number of different formats, such as traditional human-viewable stereo pairs, radial stereo pairs, and general case quantitative cylispheric stereo pairs. The QCSPs can be used for various three-dimensional imaging processes, such as passive volumetric surveying and production of three-dimensional visualization models.
I. Generation of Quantitative Cylispheric Stereo Pairs (QCSP'-s).
The first step in the surveying process is to collect image data, spatial position data and camera calibration data. After this data has been collected, two images having an overlapping scene are identified. The two identified images can come from different cameras or can be from the same camera at different points in time. The two images need not be coplanar. Each of the two identified source images is transformed into a transformed or destination image. The transformation is referred to as a “cylispheric” transformation because both cylindrical and spherical mathematical constructs are used to transform the images. The two transformed images comprise a QCSP, in that if a particular point is seen in both source images, then the point will lie on the same line in both of the transformed images. The cylispheric transform and its general case output will be described first. Quantitative radial and traditional stereo pairs are alternative output formats of the cylispheric transformation.
The cylispheric transform is most useful in the presence of accurate image parameters. Errors in the image parameters will cause the pixel associated with an object-point to have incorrect horizontal and/or vertical offsets in where the pixel is placed in the transformed image; thus, errors tend to force widening of the correlation window and hence waste computation time. The more accurate the image parameters are, the more narrow the search will be for corresponding pixels between images. In other words, with very accurate image parameters, corresponding pixels will be found in the same line of the transformed images. With less accurate positional information, corresponding pixels will be found within a few lines or more.
In what follows, “camera location” refers to where a camera was, in six-dimensional space, when a particular image was taken. The “image parameters” of each image include this information.
A first sphere 22A is centered at first camera location 14A and a second sphere 22B is centered at second camera location 14B. The angle between cylinder axis 12 and any vector originating at the origin of the sphere 22 is called “alpha”. In
There is no explicit length associated with the (alpha, beta) coordinate pair. Thus, spheres 22A and 22B of
The net result of the cylispheric transformation is that each camera location 14A and 14B is enclosed in a globe-like grid with: (1) the polar axis aligned with the cylinder axis, (2) longitude lines corresponding to rows of the output image, and (3) latitude lines corresponding to columns of the output image. The projection of the source image onto this grid illustrates where the image will plot in the output image. The cylispheric transformation does not involve any translation operations, only rotation operations. The transform is anchored to the displacement vector between the two camera locations. Thus, no parallax errors are introduced by the transformation.
The cylindrical axis vector is normalized to generate a unit vector. The cylindrical axis unit vector is then assigned to the third column of a 3×3 rotation transformation matrix. (Block 52).
The next step in the transform is to define the average camera viewing direction. (Block 54). The average camera viewing direction is calculated by summing the first column of the direction cosine matrix associated with the first image (which is, roughly speaking, the direction of the outward optical axis of the first camera) with the first column of the direction cosine matrix associated with the second image (which is, roughly speaking, the direction of the outward optical axis of the second camera), and normalizing the resulting vector.
Next, a vector is assigned to the second column of the rotation transformation matrix (Block 56); this vector is both perpendicular to the cylinder axis and perpendicular to the average viewing direction. The vector that is assigned to the third column of the rotation transformation matrix is the vector that results from the cross product of the normalized cylindrical axis vector and the normalized average viewing direction vector. The resulting vector is normalized prior to assignment to the third column of the rotation transformation matrix.
If the average viewing direction is close to the cylindrical axis, it may be preferable to use an “up” or “down” vector in place of the average viewing direction. In addition, if the cylindrical axis is vertical, it may be preferable to use a “north” vector in place of the average viewing direction.
The rotation transformation matrix is completed by assigning a vector to the first column of that matrix. (Block 58). The vector assigned to the first column of the matrix is the vector that results from the cross product of the second column of the matrix and the third column of the matrix. The completed rotation transformation matrix is used in rotating vectors expressed in global ECEF coordinates to local cylispheric X, Y and Z coordinates. Alpha and beta values may then be calculated from the cylispheric X, Y and Z coordinates.
In the following paragraphs, a “ray” is a vector which originates from a specific point and extends in only one direction (i.e., a ray is a directed line segment).
After the rotation transformation matrix is completed, the field of view for the transformed images is determined. (Block 60). Determination of the field of view involves identifying the minimum and maximum values of alpha and beta. A preferred method for identifying minimum and maximum values of alpha and beta is to choose a subset of pixels in each source image. An 11×11 grid of pixels, which includes the edges of the image has been found to be satisfactory (thus testing 121 pixels). For each pixel, a corresponding ray is generated that extends outward from the camera location. The ray is expressed in global ECEF coordinates. An alpha and beta value are calculated for each ray. In calculating the alpha and beta values that correspond to a ray, the global coordinates of the ray are transformed to cylispheric X, Y and Z coordinates using the rotation transformation matrix. The alpha and beta values are then determined from the cylispheric X, Y and Z coordinates. The alpha and beta values are compared to one another to identify the minimum and maximum values.
Central alpha and central beta values are calculated from the minimum and maximum values of alpha and beta. (Block 62). The central alpha value is calculated by adding the minimum and maximum values of alpha and dividing the result by two. Similarly, the central beta value is calculated by adding the minimum and maximum values of beta and dividing the result by two. The central alpha and central beta values will lie at the center of the transformed images which cover the full field of view of the input images.
It may be desirable to test the range of alpha values and/or beta values for usability in the traditional or radial output formats. For example, if the range of alpha values is neither all positive nor all negative, the log-scaled radial stereo images cannot be output. If the range of beta values is equal or greater than 180 degrees, then the traditional stereo images can not be output. (This is done in Block 64).
It may be desirable to further restrict the minimum and maximum values of alpha and beta to a subset of the full field of view of both images. For example, the region bounded by the limits of overlapping of the two images might be used. This is done in Block 66.
Both transformed images of the QCSP have the same direction cosine matrix.
The number of pixels to be used in each output image is defined. (Block 68). The number of pixels chosen will depend on the desired horizontal and vertical output resolution.
A step value (i.e., increment) is determined for alpha and beta. (Block 70). The step value for beta is based on the number of rows in the output image and the calculated range of beta values in the output image (i.e., maximum value of beta minus minimum value of beta), so that each row of the output image has a corresponding beta value. Similarly, the step value for alpha is based on the number of columns in the output image and the calculated range of alpha values in the output image (i.e., maximum value of alpha minus minimum value of alpha), so that each column of the output image has a corresponding alpha value.
For each pair of alpha and beta values, a corresponding ray is generated in cylispheric X, Y, Z coordinates. (Block 72). The ray is then converted to global coordinates using the rotation transformation. For each ray that is generated, the intersection of the ray with each of the source images is determined. (Block 74). The pixel in each of the source images that is intersected by the ray is copied to the corresponding destination image at the alpha and beta values which were used to generate the ray.
When identifying a ray that intersects a given pixel in the source image, or when identifying a pixel that is intersected by a given ray, camera calibration data is taken into account. To generate a ray that intersects a given pixel in the source image, the input pixel is converted into a vector in a camera coordinate system. In the camera coordinate system, the X axis points out from the center of the lens, the Y axis points to the right and the Z axis points down. The X component of the vector in camera coordinates is set to 1. The Y component is defined by multiplying the normalized horizontal pixel location by a horizontal scale factor and then adding a horizontal zero point offset. Similarly, the Z component is defined by multiplying the normalized vertical pixel location by a vertical scale factor and then adding a vertical zero point offset. The scale factors and zero point offsets are based on measured camera calibration parameters. The generated vector represents a point in the image with no distortion. Radial distortion is taken into account by first calculating the radial distance of the point from the center of the image. The radial distance is calculated by squaring the Y and Z components of the generated vector, adding the squared components, and calculating the square root of the sum. The radial distance is input to a cubic polynomial distortion correction algorithm. The cubic polynomial distortion correction algorithm outputs a distortion corrected radial distance. In a preferred embodiment, the distortion corrected radial distance is calculated by cubing the input radial distance, multiplying the cubed input radial distance by a camera specific scalar distortion factor, and adding the input radial distance. The camera specific distortion factor varies from camera to camera and depends primarilyon the amount of distortion produced by the camera lenses. Camera image planes tend to have relatively little distortion at the pixel level. Experience has shown that distortion corrections based on the radial distance from the optical axis are quit satisfactory. The single coefficient approach reduces the complexity and size of the data collection needed for camera calibration. The vector with no distortion is then adjusted for distortion by multiplying the Y and Z components of the vector by the ratio of the distortion corrected radial distance to the originally calculated radial distance. The distortion adjusted vector identifies the true point on the focal plane. The distortion adjusted vector is multiplied by the direction cosine matrix of the image to convert the vector from camera coordinates to global coordinates, resulting in a global ray.
Another situation in which camera calibration data is taken into account is when identifying a pixel in the source image that is intersected by a given ray. The process is essentially the reverse of that described above (i.e., the process of generating a ray that intersects a given pixel). However, there is one important difference. When starting with a pixel, a ray can always be generated that intersects that pixel. In contrast, when starting with a ray, that ray may or may not intersect a pixel. If a pixel is not intersected, an appropriate “fail” flag is set to so indicate.
A first step in identifying a pixel that is intersected by a given ray is to multiply the normalized ray by the in verse of the direction cosine matrix of the image to convert the ray from global coordinates to X, Y and Z camera coordinates. The Y and Z components of the input ray are each divided by the X component of the input ray to generate a vector that lies in the image plane and identifies the true point in the focal plane. The Y and Z components of the vector are used to calculate the radial distance of the true point from the center of the image. The radial distance is calculated by squaring the Y and Z components, adding the squared components, and calculating the square root of the sum. The radial distance and the camera specific distortion factor are input to a cubic polynomial equation solving algorithm that solves a cubic equation for one real root. Techniques for obtaining solutions to cubic equations are described in math textbooks such as Schaum's Math Handbook. The cubic polynomial equation solving algorithm outputs a distortion corrected radial distance. A normalized horizontal pixel location is calculated by multiplying the Y component of the image plane vector by the ratio of the distortion corrected radial distance and the original radial distance and then subtracting the ratio of the horizontal zero point offset and the horizontal scale factor. A normalized vertical pixel location is calculated by multiplying the Z component of the image plane vector by the ratio of the distortion corrected radial distance and the original radial distance and then subtracting the ratio of the vertical zero point offset and the vertical scale factor.
The cylispheric transform operates in three different modes. Calculation of the third and most general mode has been described above. Modes one and two adjust the horizontal pixel locations in useful ways. Mode one modifies the angles used in the direction cosine matrix and uses non-linear increments in the spacing between the horizontal lines of the output image. In the first mode, the transform generates a “traditional” stereo pair which can be viewed and interpreted using normal human perception. The first mode works well for images that are looking generally sideways to the vector between the image locations. The first mode does not work well for images aligned with (i.e., pointing in) the direction of travel, but the second mode works well for such images. In the second mode, the transform generates a “radial” stereo pair. The second mode does not work well if the images are not aligned with the direction of travel. The third mode is generally applicable to any pair of images that share a common scene. In the third mode, the transform generates a “general case cylispheric” stereo pair. The choice of the mode depends on the orientation of the input images, and the desired output characteristics. Examples of each of the types of QCSP'ss are discussed in the following paragraphs.
A. Traditional Stereo Pairs
Mode One QCSP'-s are modified from the general case in the following ways:
1. Note that the choice of Beta=0 (i.e. the yaw angle) of the QCSP is somewhat flexible if not arbitrary.
2. An image coordinate frame (ICF) is rolled +90 degrees with respect to the QCSP coordinate frame. (+X is forward into the image, +Y is right, +Z is down).
3. Traditional Stereo, pairs are projected onto two coplanar planes which are perpendicular to the +X axis of the ICF.
4. For projection planes at distance x=+D in the ICF, image plane point (D,Y,Z) is related to alpha and beta by:
5. The typical trapezoidal footprint of input images, and the fact that the cameras may not be looking perpendicular to the cylindrical axis, leads to an image overlap area which is centered neither on the central axis of the source images nor on the central axis of the output image coordinates. Thus, the parameters of Mode One QCSP'-s include a two dimensional offset shift from the origin of the stereo pair coordinate frame to the center of the stored synthetic image pair. The resulting benefit is that actual image size is reduced and computer storage is saved, without loss of human viewability.
B. Radial Stereo Pairs
A particularly clever, particularly useful choice of horizontal scaling around the direction of travel has been used. By taking the logarithm of the displacement from center (i.e the logarithm of the tangent of (90 degrees−ABS (alpha)), flat objects such as signs which are perpendicular to the direction of displacement of the cameras have the same size. Since most highway signs are approximately perpendicular to vehicle direction of travel, a broadly useful class of stereo correlation is thus simplified. This capability is demonstrated in
C. Cylispheric Stereo Pairs
II. Correlation of QCSP's and Passive Range Determination.
After QCSP'-s have been generated, correlation algorithms operate on the stereo pairs to identify corresponding pixels between the images. It is difficult to correlate images that are far from coplanar. For example, as one moves closer to an object, the object gets bigger. As an object changes in size from image to image, the search for corresponding pixels becomes more difficult. By making the non-coplanar images into a QCSP'-s as discussed above, the search is greatly simplified. In the QCSP'-s of the present invention, corresponding pixels between images are on the same line or within a few lines of each other. This greatly increases the efficiency of the correlation process as the search for corresponding pixels is limited to a narrow linear space. Existing correlation algorithms such as those that work with traditional stereo pairs may also be used in conjunction with the QCSP'-s of the present invention. Such algorithms identify corresponding pixels by edge detection and pattern recognition, as well as other techniques.
After corresponding pixels have been identified, the range to those pixels is determined. Using the alpha and beta values associated with each pixel, two intersecting rays in cylispheric space are computed. The intersection, or point of closest approach, of the rays are the cylispheric X,Y,Z coordinates of the point. The point can then be rotated and translated into a user-defined coordinate system. Thus, a range for the pixel may be determined by triangulation. The range determination process results in X, Y, and Z cylispheric coordinates being associated with each pixel. In a preferred embodiment, ITRF earth-centered, earth-fixed (ECEF) coordinates are used for the global X, Y, and Z coordinates. When the X, Y, Z location of a pixel has been determined, the pixel becomes a volumetric entity. In a preferred embodiment, a computer automatically correlates the pixels in a QCSP's and produces a three-dimensional visualization model.
III. Data Structure for QCSP'-s
QCSP images are preferably stored as a data structure that includes sufficient parametric information to make the images useful tools for identifying the three dimensional quantitative spatial position of elements in the scene, and for enabling the insertion of virtual (including three-dimensional) objects into the scene. The insertion of virtual objects is particularly useful in mode one, so that people can “see how things will look” in the context of the existing scene, in a natural way, after a proposed change is made. The data structure of QCSP'-s allows real-time extraction and insertion of quantitative information in three-dimensions. Typically, during the synthesis of the synthetic quantitative stereo pair, the modeled distortions in the source images are removed, so that no distortion parameters are needed to describe the output image geometry.
In a preferred embodiment, the QCSP'-s are stored as standard 24 bits/pixel bitmap images (.BMP's). The QCSP'-s include a 54 byte header, followed immediately by a continuous stream of pixel data (encoded at 3 bytes per pixel—blue/green/red). Starting at the bottom left corner of each image, the pixel data is written from left-to-right for each horizontal row of the image, with the end of the row padded with zero bytes to round up to the nearest four-byte boundary. The next row above follows immediately. In the QCSP's, one or more extra lines of black pixels at the bottom of the image provide space for quantitative header data. The quantitative header data immediately follows the standard bitmap header. The quantitative header data includes numerical coefficients that are used by a set of equations that enable the precise three-dimensional geo-positioning of corresponding pixels. The coefficients are also used in the reverse transformation from three-dimensional geo-positions to pixel-pair coordinates. The last entries in the quantitative header are a test pixel pair and associated ECEF coordinates computed using the coefficients provided.
The synthetic stereo images use two right hand Euclidean coordinate systems: (1) internal to the stereo pair, and (2) external to the stereo pair. The coordinate system external to the stereo pair uses ECEF coordinates. A rotation transformation is used to go between the internal coordinate system of the stereo pair and the external ECEF coordinates.
Although the present invention has been described with reference to preferred embodiments, workers skilled in the art will recognize that changes may be made in form and detail without departing from the spirit and scope of the invention.
|Cited Patent||Filing date||Publication date||Applicant||Title|
|US4754327||Mar 20, 1987||Jun 28, 1988||Honeywell, Inc.||Single sensor three dimensional imaging|
|US5446798 *||Feb 24, 1994||Aug 29, 1995||Fujitsu Limited||Method and apparatus for measuring position and orientation of an object based on a sequence of projected points|
|US5602584 *||Sep 27, 1995||Feb 11, 1997||Canon Kabushiki Kaisha||Apparatus for producing a panoramic image using a plurality of optical systems|
|US5633946||Nov 3, 1995||May 27, 1997||Geospan Corporation||Method and apparatus for collecting and processing visual and spatial position information from a moving platform|
|US5644651||Mar 31, 1995||Jul 1, 1997||Nec Research Institute, Inc.||Method for the estimation of rotation between two frames via epipolar search for use in a three-dimensional representation|
|US5655033||Jun 17, 1994||Aug 5, 1997||Canon Kabushiki Kaisha||Method for extracting corresponding point in plural images|
|US5680474||Oct 26, 1993||Oct 21, 1997||Canon Kabushiki Kaisha||Corresponding point extraction method for a plurality of images|
|US5703961||Dec 29, 1994||Dec 30, 1997||Worldscape L.L.C.||Image transformation and synthesis methods|
|US5937015||Sep 19, 1996||Aug 10, 1999||Dent; Paul W.||Interference mitigation by joint decoding of overlapped signals|
|US5937079||Sep 5, 1997||Aug 10, 1999||Daimler-Benz Ag||Method for stereo image object detection|
|US5963247||Dec 5, 1996||Oct 5, 1999||Banitt; Shmuel||Visual display systems and a system for producing recordings for visualization thereon and methods therefor|
|US6011863||Mar 31, 1998||Jan 4, 2000||Nec Research Institute, Inc.||Cylindrical rectification to minimize epipolar distortion|
|US6064760||Mar 1, 1999||May 16, 2000||The United States Corps Of Engineers As Represented By The Secretary Of The Army||Method for rigorous reshaping of stereo imagery with digital photogrammetric workstation|
|US6078701||May 29, 1998||Jun 20, 2000||Sarnoff Corporation||Method and apparatus for performing local to global multiframe alignment to construct mosaic images|
|US6084979||Jun 20, 1996||Jul 4, 2000||Carnegie Mellon University||Method for creating virtual reality|
|US6157747||Aug 1, 1997||Dec 5, 2000||Microsoft Corporation||3-dimensional image rotation method and apparatus for producing image mosaics|
|1||M. Pollefeys, R. Koch, M. Vergauwen and L. Van Gool, Hand-held acquisition of 3D models wilth a video camera, 1999, pp. 14-23.|
|2||Sebastien Roy, Jean Meunier, and Ingemar J. Cox, Cylindrical Rectification to Minimize Epipolar Distortion, Computer Vision and Pattern Recognition 1997, pp. 393-399.|
|Citing Patent||Filing date||Publication date||Applicant||Title|
|US7787659||Aug 6, 2008||Aug 31, 2010||Pictometry International Corp.||Method and apparatus for capturing, geolocating and measuring oblique images|
|US7873238||Aug 29, 2007||Jan 18, 2011||Pictometry International Corporation||Mosaic oblique images and methods of making and using same|
|US7991226||Oct 12, 2007||Aug 2, 2011||Pictometry International Corporation||System and process for color-balancing a series of oblique images|
|US7995799||Aug 10, 2010||Aug 9, 2011||Pictometry International Corporation||Method and apparatus for capturing geolocating and measuring oblique images|
|US8155433||Jul 10, 2008||Apr 10, 2012||Goodrich Corporation||Method of object location in airborne imagery using recursive quad space image processing|
|US8213749||Sep 17, 2010||Jul 3, 2012||Verderi, LLC||System and method for creating, storing and utilizing images of a geographic location|
|US8280107 *||Nov 5, 2007||Oct 2, 2012||Tomtom Global Content B.V.||Method and apparatus for identification and position determination of planar objects in images|
|US8385672||Apr 30, 2008||Feb 26, 2013||Pictometry International Corp.||System for detecting image abnormalities|
|US8401222||May 22, 2009||Mar 19, 2013||Pictometry International Corp.||System and process for roof measurement using aerial imagery|
|US8406513||Feb 21, 2012||Mar 26, 2013||Goodrich Corporation||Method of object location in airborne imagery using recursive quad space image processing|
|US8477190||Jul 7, 2010||Jul 2, 2013||Pictometry International Corp.||Real-time moving platform management system|
|US8497905||Sep 23, 2009||Jul 30, 2013||nearmap australia pty ltd.||Systems and methods of capturing large area images in detail including cascaded cameras and/or calibration features|
|US8520079||Feb 14, 2008||Aug 27, 2013||Pictometry International Corp.||Event multiplexer for managing the capture of images|
|US8531472||Dec 3, 2007||Sep 10, 2013||Pictometry International Corp.||Systems and methods for rapid three-dimensional modeling with real façade texture|
|US8588547||Aug 5, 2008||Nov 19, 2013||Pictometry International Corp.||Cut-line steering methods for forming a mosaic image of a geographical area|
|US8593518||Jan 31, 2008||Nov 26, 2013||Pictometry International Corp.||Computer system for continuous oblique panning|
|US8670961||Nov 2, 2011||Mar 11, 2014||Eagle View Technologies, Inc.||Aerial roof estimation systems and methods|
|US8675068||Apr 11, 2008||Mar 18, 2014||Nearmap Australia Pty Ltd||Systems and methods of capturing large area images in detail including cascaded cameras and/or calibration features|
|US8774525||Feb 1, 2013||Jul 8, 2014||Eagle View Technologies, Inc.||Systems and methods for estimation of building floor area|
|US8818138||Jun 1, 2012||Aug 26, 2014||Enrico Di Bernardo||System and method for creating, storing and utilizing images of a geographical location|
|US8818770||Apr 3, 2012||Aug 26, 2014||Eagle View Technologies, Inc.||Pitch determination systems and methods for aerial roof estimation|
|US8823732||Dec 17, 2010||Sep 2, 2014||Pictometry International Corp.||Systems and methods for processing images with edge detection and snap-to feature|
|US8825454||May 17, 2012||Sep 2, 2014||Eagle View Technologies, Inc.||Concurrent display systems and methods for aerial roof estimation|
|US9097532 *||Jan 20, 2010||Aug 4, 2015||Honeywell International Inc.||Systems and methods for monocular airborne object detection|
|US9129376||Jul 31, 2014||Sep 8, 2015||Eagle View Technologies, Inc.||Pitch determination systems and methods for aerial roof estimation|
|US9135737||Aug 1, 2014||Sep 15, 2015||Eagle View Technologies, Inc.||Concurrent display systems and methods for aerial roof estimation|
|US9183538||Mar 19, 2012||Nov 10, 2015||Pictometry International Corp.||Method and system for quick square roof reporting|
|US9262818||Jan 17, 2013||Feb 16, 2016||Pictometry International Corp.||System for detecting image abnormalities|
|US9275080||Mar 15, 2013||Mar 1, 2016||Pictometry International Corp.||System and method for early access to captured images|
|US9275496||Feb 27, 2015||Mar 1, 2016||Pictometry International Corp.||Systems and methods for rapid three-dimensional modeling with real facade texture|
|US9292913||Jan 31, 2014||Mar 22, 2016||Pictometry International Corp.||Augmented three dimensional point collection of vertical structures|
|US9330494||Oct 26, 2009||May 3, 2016||Pictometry International Corp.||Method for the automatic material classification and texture simulation for 3D models|
|US9437029||Jan 10, 2014||Sep 6, 2016||Pictometry International Corp.||Mosaic oblique images and methods of making and using same|
|US9443305||Mar 11, 2013||Sep 13, 2016||Pictometry International Corp.||Method and apparatus for capturing, geolocating and measuring oblique images|
|US9503615||Feb 26, 2015||Nov 22, 2016||Pictometry International Corp.||System and process for color-balancing a series of oblique images|
|US9514568||Mar 3, 2014||Dec 6, 2016||Eagle View Technologies, Inc.||Aerial roof estimation systems and methods|
|US9520000||Feb 29, 2016||Dec 13, 2016||Pictometry International Corp.||Systems and methods for rapid three-dimensional modeling with real facade texture|
|US9542738||Mar 3, 2016||Jan 10, 2017||Pictometry International Corp.||Augmented three dimensional point collection of vertical structures|
|US9599466||Feb 1, 2013||Mar 21, 2017||Eagle View Technologies, Inc.||Systems and methods for estimation of building wall area|
|US9612598||Jan 7, 2015||Apr 4, 2017||Pictometry International Corp.||Unmanned aircraft structure evaluation system and method|
|US9633425||Feb 12, 2016||Apr 25, 2017||Pictometry International Corp.||System for detecting image abnormalities|
|US9644968||Aug 11, 2014||May 9, 2017||Vederi, Llc||System and method for creating, storing and utilizing images of a geographical location|
|US9652980||Apr 23, 2014||May 16, 2017||GM Global Technology Operations LLC||Enhanced clear path detection in the presence of traffic infrastructure indicator|
|US9753950||Mar 15, 2013||Sep 5, 2017||Pictometry International Corp.||Virtual property reporting for automatic structure detection|
|US20080123994 *||Aug 29, 2007||May 29, 2008||Stephen Schultz||Mosaic Oblique Images and Methods of Making and Using Same|
|US20080204570 *||Feb 14, 2008||Aug 28, 2008||Stephen Schultz||Event Multiplexer For Managing The Capture of Images|
|US20080231700 *||Jan 31, 2008||Sep 25, 2008||Stephen Schultz||Computer System for Continuous Oblique Panning|
|US20080273753 *||Apr 30, 2008||Nov 6, 2008||Frank Giuffrida||System for Detecting Image Abnormalities|
|US20090096884 *||Aug 6, 2008||Apr 16, 2009||Schultz Stephen L||Method and Apparatus for Capturing, Geolocating and Measuring Oblique Images|
|US20090097744 *||Oct 12, 2007||Apr 16, 2009||Stephen Schultz||System and Process for Color-Balancing a Series of Oblique Images|
|US20090141020 *||Dec 3, 2007||Jun 4, 2009||Freund Joseph G||Systems and methods for rapid three-dimensional modeling with real facade texture|
|US20090256909 *||Apr 11, 2008||Oct 15, 2009||Nixon Stuart||Systems and methods of capturing large area images in detail including cascaded cameras and/or calibration features|
|US20100008565 *||Jul 10, 2008||Jan 14, 2010||Recon/Optical, Inc.||Method of object location in airborne imagery using recursive quad space image processing|
|US20100013927 *||Sep 23, 2009||Jan 21, 2010||Nearmap Pty Ltd.||Systems and Methods of Capturing Large Area Images in Detail Including Cascaded Cameras and/or Calibration Features|
|US20100166256 *||Nov 5, 2007||Jul 1, 2010||Marcin Michal Kmiecik||Method and apparatus for identification and position determination of planar objects in images|
|US20100296693 *||May 22, 2009||Nov 25, 2010||Thornberry Dale R||System and process for roof measurement using aerial imagery|
|US20100302243 *||Aug 10, 2010||Dec 2, 2010||Schultz Stephen L||Method and apparatus for capturing geolocating and measuring oblique images|
|US20110026772 *||Apr 18, 2008||Feb 3, 2011||Tele Atlas B.V.||Method of using laser scanned point clouds to create selective compression masks|
|US20110096083 *||Oct 26, 2009||Apr 28, 2011||Stephen Schultz||Method for the automatic material classification and texture simulation for 3d models|
|US20110178658 *||Jan 20, 2010||Jul 21, 2011||Honeywell International Inc.||Systems and methods for monocular airborne object detection|
|US20110187713 *||Feb 1, 2011||Aug 4, 2011||Eagle View Technologies, Inc.||Geometric correction of rough wireframe models derived from photographs|
|US20150245020 *||Jun 5, 2013||Aug 27, 2015||Metaio Gmbh||Image processing method, particularly used in a vision-based localization of a device|
|U.S. Classification||382/154, 382/293|
|International Classification||G01C11/06, G06T1/00, G06T7/60, G06K9/32, G06T7/00, G06K9/00|
|Jun 20, 2002||AS||Assignment|
Owner name: GEOSPAN CORPORATION, MINNESOTA
Free format text: ASSIGNMENT OF ASSIGNORS INTEREST;ASSIGNOR:SETTERHOLM, JEFFREY M.;REEL/FRAME:013311/0854
Effective date: 20020613
|Oct 23, 2007||CC||Certificate of correction|
|Jan 24, 2011||REMI||Maintenance fee reminder mailed|
|Apr 29, 2011||FPAY||Fee payment|
Year of fee payment: 4
|Apr 29, 2011||SULP||Surcharge for late payment|
|Mar 10, 2014||AS||Assignment|
Owner name: TORETTA, PAUL A., MINNESOTA
Free format text: NOTICE OF JUDGEMENTS;ASSIGNOR:GEOSPAN CORPORATION;REEL/FRAME:032613/0581
Effective date: 20040304
|Mar 21, 2014||AS||Assignment|
Owner name: TORETTA, PAUL A., MINNESOTA
Free format text: NOTICE OF JUDGEMENT;ASSIGNOR:GEOSPAN CORPORATION;REEL/FRAME:032613/0591
Effective date: 20140313
|Nov 26, 2014||FPAY||Fee payment|
Year of fee payment: 8